CUDA

Question 1

What is heterogeneous computing?

Answer 1

Computing in both GPU and CPU

Question 2

What is the Host?

Answer 2

CPU and its memory

Question 3

What is the Device?

Answer 3

GPU and its memory

Question 4

How do we declare a kernel function that is to be run on the Device?

Answer 4

__global__

Question 5

What does the keyword __global__ indicate?

Answer 5

A function that runs on the device and is called from host code.

Question 6

What does it mean when we describe launching a kernel?

Answer 6

The CPU places the kernel into the GPGPU stream. Execution in the CPU continues without waiting for the kernel execution to complete.

Question 7

How does the CPU launch a kernel on the device?

Answer 7

Triple angle brackets mark a call from the host to the device.

Question 8

Device Pointers

Answer 8

Points to GPU memory.
May to passed to/from host code.
May not be dereferenced in host code.

Question 9

Host Pointers

Answer 9

Point to CPU memory.
May be passed to/from device code.
May not be dereferenced in device code.

Question 10

What is used to allocate memory on the GPGPU device?

Answer 10

cudaMalloc()

Question 11

What is used to copy memory from the CPU Host to the GPGPU?

Answer 11

cudaMemcpy() with the cudamemcpyHostToDevice.

Question 12

What is used to copy memory from the GPGPU to the CPU?

Answer 12

cudaMemcpy() with the cudamemcpyDeviceToHost.

Question 13

Where are Device pointers stored?

Answer 13

Stored on the Host and passed to the kernels when they execute.

Question 14

What is a thread block?

Answer 14

A group of threads executed together.

Question 15

How are threads in different thread blocks synchronized?

Answer 15

Call another kernel.

Question 16

How can threads within a thread block synchronize?

Answer 16

__syncthreads();

Question 17

What threads can access the memory that is declared using the __shared__ keyword?

Answer 17

Threads in the same block.

Question 18

What purpose does the variable blockIdx.x have in CUDA?

Answer 18

Access block index.

Question 19

What purpose does the variable threadIdx.x have in CUDA?

Answer 19

Access thread index within block.

Question 20

What purpose does the variable blockDim.x have in CUDA?

Answer 20

Get threads per block

Question 21

For a kernel launch two parameter are given inside of the «<»>, what is the prupse of these parameters?

Answer 21

«<nblocks, nThread per block»>

Question 22

What does __syncthreads() do? If a GPGPU executes in a SIMD mode, why do we need this call?

Answer 22

Used as a barrier to prevent data hazards.

Threads are divided into small groups of threads called warps. While the warps execute in a SIMD manner and are implicitly synchronized, the threads in a thread block may not remain synchronized.